About ScrubBeforeAI
ScrubBeforeAI was born out of a daily engineering struggle: the tension between developer velocity and data security.
As engineers, we copy-paste error logs, configuration files, and stack traces into AI assistants (like ChatGPT, Claude, and Gemini) multiple times a day to speed up debugging. However, these codeblocks frequently contain hardcoded secrets, connection strings, private hostnames, or customer PII.
Manually scanning and redacting this text takes valuable time, leading many to take the risk and paste raw logs. When developers paste sensitive credentials, they risk leaking them into AI training sets or exposing them to third-party data tracking.
To build a zero-trust, developer-friendly interface that removes all friction from prompt sanitization. We wanted a tool that not only redacts but also lets you restore those values in the AI output with a single click — all processed completely client-side in the browser.
The Problem We Solve
Every day, millions of developers around the world paste code snippets, error logs, and configuration files into generative AI assistants. These prompts often contain sensitive information that should never leave a developer's local machine: API keys for cloud services like AWS, Google Cloud, and Azure; database connection strings with embedded passwords; OAuth tokens and JWT secrets; personally identifiable information such as email addresses, phone numbers, and IP addresses; and internal infrastructure hostnames and network endpoints.
Most AI platforms retain prompt history for model training, quality assurance, or audit purposes. Once a credential is submitted in a prompt, it is permanently stored on external servers outside your control. If the AI provider experiences a data breach, or if the model inadvertently memorizes and reproduces a credential to another user, your infrastructure is compromised.
ScrubBeforeAI exists to close this gap. By providing an instant, client-side boundary between your raw data and AI chat interfaces, we ensure that sensitive values are replaced with safe placeholders before they ever leave your browser.
Our Core Principles
100% Client-Side Privacy
All parsing, pattern-matching, and text substitutions occur directly in your browser. We run no database, no backend server processing, and zero analytical trackers beyond anonymous page-view statistics. Your code never leaves your local workspace. The entire redaction engine runs inside your browser's JavaScript sandbox, meaning your raw credentials are never transmitted over any network connection.
Two-Way Redaction
We believe security tools should improve your workflow, not break it. By keeping a local reversal session map, you can paste the AI's response back to reconstruct the original credentials instantly. This two-way workflow means you get the full benefit of AI-assisted debugging without ever compromising your secrets. The session map is stored in temporary browser memory and is destroyed when you close the tab.
Instant Sandbox Execution
No signups, user accounts, or extensions required. It runs instantly inside a clean, sandboxed local session context, keeping your credentials isolated from external AI tracking. You can start using ScrubBeforeAI within seconds of opening the page — no installation, no configuration, no browser extensions to manage.
Offline-First Architecture
ScrubBeforeAI is built as a Progressive Web App (PWA) with a cache-first service worker. After your initial visit, the entire application — including all redaction patterns, the reversal engine, the file scanner, and the custom rules interface — works completely offline. This makes it ideal for air-gapped environments, secure corporate networks, and scenarios where internet access is restricted or unreliable.
Who We Serve
ScrubBeforeAI is designed for anyone who works with sensitive data and uses AI assistants as part of their daily workflow. Our primary users include:
- Software Developers: Backend and full-stack engineers who debug production issues by pasting stack traces, error logs, and environment configurations into ChatGPT, Claude, or Gemini.
- DevOps & SRE Teams: Infrastructure engineers who work with deployment scripts, CI/CD pipeline logs, and cloud provider credentials on a daily basis.
- Security Engineers: Professionals who need to analyze log files and incident reports without exposing sensitive indicators of compromise to external systems.
- Data Engineers: Teams working with database connection strings, data pipeline configurations, and ETL scripts that contain embedded credentials.
- Compliance Officers: Professionals responsible for ensuring that engineering teams adhere to GDPR, SOC2, HIPAA, and PCI-DSS data handling requirements when using AI tools.
Technology Stack
ScrubBeforeAI is built with modern, performance-optimized web technologies chosen specifically for client-side security and speed:
- Astro: A modern static site generator that pre-renders all pages at build time, ensuring fast load times and excellent SEO. Astro's islands architecture allows us to hydrate only the interactive components, keeping the page lightweight.
- React: Powers the interactive redaction workspace, file scanner, and custom rules engine. React's component model lets us build a responsive, real-time interface that updates instantly as you type.
- Client-Side Regex Engine: Our detection system uses over 40 priority-sorted regular expression patterns running natively in your browser's JavaScript engine. These patterns cover cloud provider keys, database URIs, tokens, PII formats, and more.
- Web Crypto API: Session maps can be exported with AES-GCM 256-bit encryption using the browser's native Web Crypto API, ensuring your reversal data is protected even when saved to disk.
- Service Worker (PWA): A cache-first service worker enables full offline functionality after the initial page load.
Contact Us
Have feedback, found a bug, or want to suggest new regex patterns? We would love to hear from you. Visit our Contact page to send us a message directly and securely.